Jinning Li

Research Intern at Cornell University

Counterfactual Machine Learning

Advisor: Prof. Thorsten Joachims

Summer 2018 ~ Fall 2018

Ad Placement based on Criteo dataset

Using logged bandit feedback to learn an Ad placement policy. Got Rank 1 (post-challenge) in Criteo Ad Placement Challenge.

A Hybrid Framework of Counterfactual Risk Minimization and Supervised Learning on Logged Bandit Feedback

[Presentation] [Writeup] [Weekly Report] Propose a hybrid framework combining the advantages of CRM and supervised learning. Both the difference between the observed feedbacks and predictions, and policy propensity are optimized.

Advanced Network Laboratory at SJTU

Data Mining, Social Networks, Time Series Analysis, Machine Learning

Advisor: Prof. Xiaofeng Gao

Summer 2017 ~ Present

Cross-Platform Event Popularity Analysis

[Website] [Paper] Designing a universal scheme analyzing event popularity on Weibo and Baidu platform. Paper published in International Conference on Database and Expert Systems Applications (DEXA), 2018.

Sentiment-Aware Topic Popularity Prediction on Short Text based Social Media

Propose a novel framework, Senti2Pop, to better predict topic popularity utilizing sentiment factors. Paper submitted to SDM 2019.

Topic Detection and Dissemination Trend Analysis on Social Network

Detect and analyze trending topics on social networks with neural networks. Paper submitted to SDM 2019.

Scribble-to-Painting Transformation with Multi-task Generative Adversarial Networks (Under Reviewing)

Jinning Li, Yexiang Xue. Scribble-to-Painting Transformation with Multi-task Generative Adversarial Networks. Submission to AAAI, 2019.

Senti2Pop: Sentiment-Aware Topic Popularity Prediction on Short Text based Social Media (Under Reviewing)

Jinning Li, Qiang Zhang, Jiayi Xu, Xiaofeng Gao, Guihai Chen. Senti2Pop: Sentiment-Aware Topic Popularity Prediction on Short Text based Social Media. Submission to SDM, 2019.

DANCINGLINES: An Analytical Scheme to Depict Cross-Platform Event Popularity

[Website][Paper] Tianxiang Gao, Weiming Bao, Jinning Li, Xiaofeng Gao, Boyuan Kong, Yan Tang, Guihai Chen, Xuan Li. DancingLines: An Analytical Scheme to Depict Cross-Platform Event Popularity. International Conference on Database and Expert Systems Applications (DEXA), 2018.

Topic Detection and Dissemination Trend Analysis on Social Network (Under Reviewing)

Jiadong Chen, Tianxiang Gao, Xiaofeng Gao, Peng He, Jinning Li, Guihai Chen. Topic Detection and Dissemination Trend Analysis on Social Network. Submission to SDM, 2019.

C++ Programming

Teaching Assistant at Shanghai Jiao Tong University

https://acm.sjtu.edu.cn/wiki/Programming_2016

Fall 2016

• * Design homework and test to help students master the basic idea of C++ programming.
• * Give a lecture on dynamic programming [slide].
• * Give a lecture on final review [slide].

Criteo Ad placement Challenge

https://github.com/jinningli/ad-placement-pytorch

July. 2018

Ad placement with counterfactual machine learning. Implemented with Pytorch.
Got Rank 1 (post-challenge) in NIPS '17 Workshop: Criteo Ad Placement Challenge

DeepWave: Learning to Simulate Water Wave in Real-time

https://github.com/jinningli/DeepWave

May. 2018

[Paper][Presentation] Course project in CS230 Virtual Reality. A novel water wave simulation technique based on deep learning and wave packet theory.

Dancinglines

https://github.com/jinningli/dancinglines

July. 2018

Source Code of "DANCINGLINES: An Analytical Scheme to Depict Cross-Platform Event Popularity". Short-text based cross-platform event popularity analysis.

Compiler-2017

https://github.com/jinningli/Complier-2017

May. 2017

A compiler written in Java translating Mx* Language, a hybrid of C and Java, into X86 Assembly.

Convolutional BiMPM for Natural Language Inference

https://github.com/jinningli/CBiMPM

June. 2018

[Report] Course project in CS229 Natural Language Processing. I Propose a novel convolutional bilateral multi-perspective matching model for natural language inference task on SNLI dataset, improving the accuracy to 86.7%.

Audio Tagging

https://github.com/jinningli/Audio-Tagging

Jan. 2018

[Report] Project of MS318, Deep Learning. LSTM model and gated CNN model are used to solve the audio utterance tagging task.

Text Classification

https://github.com/jinningli/Text-Classification

Mar. 2017

[Report] A course work in CS420: Text Classification using Linear Regularized Regression Method. Using Jieba word parsing and efficient feature selecting algorithm.

Ray Tracer

https://github.com/jinningli/Ray-Tracer

Jun. 2016

[Demo] A C++ program to operate Ray Tracing on a 3D model and output a 2D picture.

A simple CPU

https://github.com/jinningli/CPU

Apr. 2017

[Report(CHS)]A course work in Computer Architecture. It's a CPU with five assembly line in Verilog language. And It ran successfully in FPGA.

MIPS Simulator

https://github.com/jinningli/Mips-Simulator

Jul. 2016

A C++ program to simulate MIPS assembly code and give the result, with virtual CPU, Register, RAM and ROM.

Text Compressor

https://github.com/jinningli/Text-Compressor

Aug. 2016

A efficient C++ program to compress text by about 50%, using Huffman Tree and Word Reduce algorithm.